This study summarizes my contributions as a product designer and thought leader to Oracle’s generative AI solutions. This was part of a spontaneous, generative exercise, and not a defined project: I worked with many designers and product managers to push the strategy and concepts of what GenAI could look like in Oracle Cloud Infrastructure's console.
During the ideation process I was part of a small group of leaders and thinkers tasked with brainstorming new ways to use an AI-assisted cloud platform: where would users want assistance? How could we leverage AI to facilitate their day-to-day? For the chatbot design, I helped lead thinking and design concepts for the AI chatbot, including evocation standards and multi-workflow user journeys. These efforts underwent many iterations of design revisions, based both off evolving needs and user research and validation.
Goal: The goal was to develop intuitive and efficient conversational interfaces that could assist users in managing their cloud resources and performing various tasks around the console, such as intelligent resource retrieval, customized networking suggestions, and more. An assistant was conceptualized as a console-wide functionality that could be called upon in any workflow, and when a user needed more in-depth help they could be guided to a full AI chatbot.
Challenges: OCI's console is already dense with information; adding an AI assistant was difficult not only regarding the question of how and when a user would want to interact with one while not getting in the way of the user's original intent, but also how to introduce a persistent functionality that would work for all OCI templates and workflows. There was the additional complication of working so loosely with such a large group, and refining rather ephemeral concepts into tangible interfaces.
Deliverables: concept proposals, design thinking and direction, design mockups.
As generative AI took the world by storm and the industry realized it’s utility, it became apparent that our users would expect the same capabilities for their infrastructure and general cloud needs. So Oracle launched the effort to offer a generative AI assistant that would ease a customer’s workflow, and I was part of the team tasked with innovating where, what, why, and how.
But AI was already being shoehorned into every available and ill-suited cranny, and my team knew we needed to be careful of not doing the same: we wanted to design an AI experience that was thoroughly designed and genuinely needed. Instead of just keeping up with competitors, we wanted to think at a grand scale and innovate how AI could be used. There were many constraints and considerations. First, the OCI console was the first place in Oracle that would implement anything like our generative AI, meaning that we had to design at scale for unknowns and future implementations. Our work would be the standard that would be upkept throughout the console. Second, data showed that users would prefer no help at all rather than something obtrusive or unhelpful (see the User Research section). We couldn’t design the next Clippy. We had to make high-level strategic decisions about how and when a user would evoke an assistant, and get out of a user’s way otherwise.
To begin, we analyzed the cause and context for the ask of an AI assistant and chatbot. Then we reviewed what existing research and standards could tell us. The reason for the ask was the industry boom that came with the introduction of ChatGPT and the growing standard for cloud platforms to integrate a similar functionality. In the spirit of being competitive but also seeing an opportunity to provide much-needed user guidance, OCI needed to introduce their own AI.
To ensure we had a solid foundation the team conducted a competitor analysis. We looked at competitors such as Cloudcraft, AWS, Thousand Eyes, and more to review their AI implementations, identifying key features, strengths, and weaknesses. This helped us benchmark our design against industry standards and discover opportunities for innovation. We categorized the typical use cases for an AI assistant, and in reverse, which areas of the cloud console would benefit from surfacing additional guidance to a user.
As a group we decided the core tenets of our AI assistant would be: guide users, unify the console, mazimize automation, minimize service requests, and allow personalization.
We focused on particular use cases to highlight how to integrate an assistant, including: resource creation guidance in Compute Instances, solution catalogs for cloud strategies, network monitoring, and more.
Other members of the design team drafted rough user journeys for the above mentioned use cases that kicked off general designs, with all designers including myself meeting together to share feedback and dispersing to ideate separately. We went through several cycles of general brainstorming and proposals by all the designers. We met weekly to review progress and findings, and to share the latest idea with leadership.
The complexity of a cloud console meant that there were many opportunities for where to evoke our assistant. One thing that became more and more apparent as we worked through designs was that evokement itself would have to be carefully considered: When or how to trigger AI in the first place? Could it be turned on or off? How would we handle the transition from simple to complex guidance and longer conversational AI? How much of what a user did or requested in AI chatbot should in turn be reflected dynamically in the console? Our designs explored all this in many different directions.
The review and revision process happened iteratively over the course of the entire project until a general direction was collaboratively refined and sent for leadership approval. At this stage my role as a high-level thinker and low-fidelity concept designer was complete; the final high-fidelity design work was assigned to a dedicated team, so we handed off the vision and materials. The AI assistant and chatbot is now a permanent part of OCI, surfacing in various workflows and components, thanks to the collaboration of myself and my teammates.
Our AI assistant concept ideation and design generation enhanced the user experience of cloud console workflows. This was achieved with a lot of collaboration, repetition, and revisions. While working in such a large team and directing such ephemeral concepts was at times frustrating, it ultimately led to in-depth and thorough solutions that not one member could have put together by themselves.
